GitHub - jgardo/lambda-for-vps: Simple service forwarding request to docker containers. If there is a request to container and container is down, it's starting it on demand

Simple service forwarding request to docker containers. If there is a request to container and container is down, it's starting it on demand. After serving request it's scheduled to be killed after some time of idle.

Requirements

This service requires docker and docker-compose. It uses local docker instance and it's local /var/run/docker.sock to administrate containers.

It depends on portainer and openresty (nginx)

User guide

Lambda-for-vps (shortcut L4v) starts by executing command

docker-compose -f docker-compose.yml up

in root of project.

By default services managed by L4v are lazy loaded, so on startup there are no additional actions.

On first request to service there are few steps to do:

1. (if necessary) Initialize local environment to manage docker.

   L4v uses portainer to manage containers. So firstly L4v initializes admin using environments portainerLogin which is default admin and portainerPassword with default adminpassword. It also creates local "portainer endpoint". If you want to check what's inside portainer it's available on port 9000.

2. (if necessary) Deploy given docker-compose.yml configured for this request.

3. Making sure that requested endpoint is up. If it's not, then it restarts all containers within docker-compose.yml configured for this request.

4. (if necessary) Wait configured amount of seconds, so all containers can be started and initialzed.

5. Serving request.

6. Schedule killing all services after configured amount of seconds. If there are another request services will be killed amount of seconds after last request.

7. If time is up, there are sended docker stop to all containers within docker-compose.yml.

Developer guide

By default there is an example of configuration for simple hello-world docker instance.

After startup example hello-world is available at http://localhost:80/hello-world. It can take a while because docker image have to be downloaded, extracted and launched. However next request should take less time.

Docker configuration is placed in file config/app/hello-world/docker-compose.yml.

L4v is based on OpenResty project, which is a kind of wrapper for nginx. So configuration of endpoints are given in file config/nginx/conf.d/default.conf.

Example hello-world application has such entry in default.conf

location /hello-world {
    # MIME type determined by default_type:
    default_type 'text/plain';
    set $ip "";
    set $port "8080";
    access_by_lua_file /hello-world/hello-world-app.lua;
    proxy_pass http://$ip:$port/;
}

Setting ip are crucial, because ip depends on inner docker container ip and is evaluated during startup.

Whole administration of request is in line access_by_lua_file /hello-world/hello-world-app.lua;. It's execution maintenance of whole docker-compose life.

By default such maintenance looks like that:

local initializer = require('initializer')
local killer = require('killer')

local initParams = {
    stackName = "hello", -- informative name of app used in Portainer
    afterRestartSleep = 2, -- constant time in seconds to wait for 
                    -- initialization of container after starting it
    killAfter = 15, -- time in seconds to kill all containers in docker-compose 
                    -- after last request
    dockerComposeFilePath = "/hello-world/docker-compose.yml" 
                    -- path to docker-compose.yml file to be deployed 
}
initializer.init(initParams);
initializer.updateTimeToKill(initParams);

killer.scheduleKilling(initParams)

Crucial part of this snippet is definition of initParams. Rest may remain unchanged.